Folate and prostate cancer a case-control study

Transcription

1 Oregon Health & Science University OHSU Digital Commons Scholar Archive May 2009 Folate and prostate cancer a case-control study Rachel Linz Follow this and additional works at: Recommended Citation Linz, Rachel, "Folate and prostate cancer a case-control study" (2009). Scholar Archive This Thesis is brought to you for free and open access by OHSU Digital Commons. It has been accepted for inclusion in Scholar Archive by an authorized administrator of OHSU Digital Commons. For more information, please contact champieu@ohsu.edu.

2 Folate and Prostate Cancer: A Case Control Study By Rachel Linz A THESIS Presented to the Department of Public Health and Preventive medicine and the Oregon Health and Science University School of Medicine in partial fulfillment of the requirements for the degree of Master of Public Health May 2009

3 Department of Public Health and Preventive Medicine School of Medicine Oregon Health & Science University CERTIFICATE OF APPROVAL This is to certify that the Master s thesis of Rachel Linz has been approved Jackilen Shannon, PhD Mark Garzotto, MD Dawn Peters, PhD

4 Table of Contents List of Tables and Figures....ii Acknowledgments..iii Abstract iv Background and Significance Specific Aims....5 Methods..6 Overview.. 6 Recruitment and Informed Consent...6 Measurement of Predictor Variables..8 Measurement of Outcome Variable..10 Statistical Analysis..10 Results...13 Demographics...13 Folate..13 Logistic Regression Modeling.20 Prostate Cancer Cases Compared to Clinic Controls..22 Prostate Cancer Cases Compared to Biopsy Negative Controls.24 Stratification by Gleason Score..28 Discussion 31 Blood Folic Acid Concentration. 31 Comparison of Cases and Clinic Controls.. 33 Stratification by Severity of Disease. 34 Comparison of Cases and Biopsy Negative Controls. 36 Strengths and Limitations.38 Future Studies Conclusion...39 References..41 i

5 List of Tables and Figures Table 1. Summary of studies examining the association between folate and PCa. 4 Table 2. Characteristics of Subjects Who Did and Did Not Provide Blood Samples. 15 Table 3. Frequencies of Demographic and Risk Factor Characteristics Table 4. Distribution of Measures of Folate Status Table 5. Correlation coefficients and corresponding p values for measures of folate status.. 19 Table 6. Univariate comparisons of dietary data between PCa cases and both control groups.21 Table 7. Additional univariate comparisons between PCa cases and two control groups. 22 Table 8. Final multivariate model comparing prostate cancer cases with clinic controls..24 Table 9. Initial multivariate model comparing prostate cancer cases with biopsy negative controls..25 Table 10. Final multivariate model comparing prostate cancer cases with biopsy negative controls Table 11. Multivariate model comparing severe prostate cancer cases (Gleason 7, n=60) with clinic controls (n=181) Table 12. Multivariate model comparing less severe prostate cancer cases (Gleason < 7, n=57) with clinic controls (n=181) Figure 1. Sample Population....8 Figure 2. Distribution of blood folic acid concentration among prostate cancer cases 18 Figure 3. Distribution of blood folic acid concentration among biopsy negative controls...18 Figure 4. Distribution of blood folic acid concentration among clinic controls..19 ii

6 Acknowledgments I would first like to thank my committee members, Dr. Jackie Shannon for inviting me to conduct the blood folic analysis for my internship and for making this project possible, Dr. Mark Garzotto for being a key member of the study team and for providing valuable clinical expertise, and Dr. Dawn Peters for sharing her extensive statistical knowledge and for keeping me on track. I would also like to thank the OHSU Department of Public Health and Preventive Medicine, especially Dr. Katie Riley and Tree Triano, for all of their help over the past two years. Finally, I would like to thank my family, friends, and classmates for all of their support. iii

7 Abstract Background: Prostate cancer is the most prevalent cancer in men in the United States. It is estimated that in the US 186,000 men will be diagnosed with prostate cancer in 2008 and that the lifetime probability of being diagnosed with prostate cancer is 16.7%. Of the currently known risk factors for prostate cancer, such increasing age, race and family history, few are modifiable. Due to geographic variation in prostate cancer incidence, researchers have been investigating environmental factors, such as diet, that may play a role in prostate cancer risk. Folate is one such dietary factor that has been thoroughly investigated with regard to colorectal cancer due to its role in one carbon metabolic pathways. However, studies investigating the role of folate in prostate cancer risk have reported conflicting results. Methods: We conducted a case control study of men with biopsy confirmed prostate cancer, men with negative biopsies, and a third group of clinic controls with no history of prostate abnormalities. Study subjects completed an interviewer administered, validated, diet history questionnaire prior to biopsy (or primary care appointment), as well as providing demographic and risk factor information and a pre biopsy blood specimen. Blood specimens were analyzed for folic acid content using an automated chemiluminescence assay, and the association between folic acid concentration and reported intake of dietary and synthetic folate and folic acid was examined. In addition, multiple logistic regression models were built to examine the significance of folic acid concentration on prostate cancer risk when combined with known risk factors. iv

8 Results: Blood folic acid concentration was not significantly associated with prostate cancer. Correlations between blood folic acid concentration and reported dietary intake of folate or folic acid were low. Conclusion: Although others have reported a significant association between folate and prostate cancer, we did not observe such an association. The timing of exposure to folate may affect prostate cancer risk. Further longitudinal studies examining folate intake are recommended. v

9 Background & Significance Prostate cancer (PCa) is the most prevalent cancer in men in the United States. It is estimated that in the US 186,000 men will be diagnosed with prostate cancer in 2008 and that the lifetime probability of being diagnosed with prostate cancer is 16.7% 11. Of the currently known risk factors for prostate cancer, such as age, race and family history, few are modifiable. Incidence of invasive prostate cancer is highest in Europe, North America and Australia, whereas the lowest incidence is found in Japan and other Asian countries 2. Results from migration studies have demonstrated that the risk of developing prostate cancer increases when men move from a low risk country to a high risk country, suggesting that environmental factors such as lifestyle and dietary habits are associated with the risk of developing invasive prostate cancer 3. Thus, recent epidemiologic studies have focused on assessing potential modifiable risk factors for prostate cancer, such as diet. One dietary component that has been implicated in the pathogenesis of several cancers is folate. Folate is an essential B vitamin found in leafy green vegetables, whole grains, and some fruits. Folate in this form is known as 5 methyltetrahydrofolate (5 MTHF) or pteroylpolyglutamate. Starting in 1998, the United States Food and Drug Administration mandated that enriched flours and cereals be fortified with folic acid (140µg/100gm serving), the fully oxidized synthetic form of folate, in order to reduce the incidence of neural tube defects (NTDs) 4. Folic acid, or pteroylmonoglutamate, is also the form of folate found in supplements. Folic acid fortification has resulted in an 1

10 increased intake among adults in the United States 5. Throughout this paper folate will refer to 5 MTHF and folic acid will refer to pteroylmonoglutamate. Folate is an essential factor in the one carbon metabolic pathway that is involved in DNA methylation, synthesis, and repair. Insufficient levels of folate can compromise nucleotide synthesis and lead to misincorporation of uracil instead of thymidine in DNA, causing DNA instability and increased mutations and chromosomal breaks 6. In several studies of colorectal cancer, the most frequently studied cancer in relation to onecarbon metabolism, high folate status has been reported to be associated with a modestly reduced risk for cancer development in prospective studies 7. However, the first randomized controlled trial investigating the potential antineoplastic effect of supplementary folic acid in subjects with a history of colorectal adenomas found an increased risk of advanced lesions in the supplementation group compared to the placebo group 8. This unexpected result suggested that folic acid plays a dual role in the neoplastic process depending on the timing of exposure. In addition to its role in nucleotide synthesis, folate is directly involved in the formation of S adenosylmethionine (SAM), which provides methyl groups for DNA methylation, a common method of gene regulation. DNA methylation involves the covalent bonding of a methyl group to a cytosine that precedes a guanosine in the DNA sequence (a CpG dinucleotide). CpG dinucleotides are uncommon in the genome as a whole, but are often clustered in small stretches of DNA called CpG islands. These islands are often found in the promoter regions of genes. About 80% of the CpG dinucleotides that are not associated with CpG islands are methylated, whereas CpG 2

11 islands, especially those found in promoter regions, are usually unmethylated. Methylation of CpG islands in promoter regions reduces or silences gene expression 9. Hypermethylation of certain genes has been observed in cancer cells and tissues for more than 20 years. Recent improvements in detection of methylation through methylation specific PCR (MSPCR) have increased the body of knowledge on the association between gene methylation status and cancer 10. Hypermethylation has been reported to occur in genes responsible for cell cycle regulation (RASSF1a, p16/cdkn2a), DNA repair (GSTP1, MGMT), and tumor growth and progression (APC, RARB2) and has been seen in several types of cancer In prostate cancer patients, promoter hypermethylation of certain genes has been shown to provide significant value in diagnosis and prognosis 14,15. Due to the multiple roles of folate in one carbon metabolic pathways, it is not surprising that studies examining the association between folate and prostate cancer have reported conflicting results. Although some studies have suggested that high folate status may increase risk of prostate cancer 16,17, these associations were not statistically significant. Others have reported a protective effect of folate and still others have reported no association at all 21,22. The Aspirin/Folate Polyp Prevention Study 23, the first randomized controlled trial of folic acid supplementation that assessed prostate cancer as an endpoint, found that although baseline dietary intake of folate was inversely associated with prostate cancer, folic acid supplementation was associated with an increased risk of prostate cancer. These intriguing results support the need for further 3

12 study to elucidate the relationship between folate intake and prostate cancer. A summary of these studies is provided in Table 1. There are several potential reasons for the discrepancy in reported findings. First, the questionnaires used assess folate intake in different ways and thus provide different estimates of folate intake. Second, those studies using serum and blood samples used a variety of methods that cannot be reliably compared 24. There is no currently accepted gold standard for folate analysis. Table 1. Summary of studies examining the association between folate and PCa First Author and Location Study Design Method of Folate Results Year Analysis Vlajinac 1997 Serbia Case control Diet questionnaire Null Weinstein 2003 Finland Case cohort Diet questionnaire; Null serum sample Hultdin 2005 Sweden Nested case control Serum sample Positive (not significant after adjustment for B12) Pelucchi 2005 Italy Case control Diet questionnaire Negative Rossi 2006 Australia Prospective cohort Serum sample; Negative blood sample Stevens 2006 Europe Prospective cohort Diet questionnaire Negative (not significant) Johansson 2008 United States Nested case control Blood sample Positive (not significant) Figueiredo 2009 United States Randomized Controlled Trial Serum sample; Diet questionnaire Negative for dietary folate; positive for folate supplementation Significance: Prostate cancer is a source of significant health burden in men in the United States. Identification of modifiable risk factors is essential to reducing the incidence of this disease. Due to increased folic acid intake since the mandatory fortification with folic acid of cereals and flours in 1998, it is critical to determine whether this increased folic acid intake is in fact advisable to all members of the US 4

13 population. In contrast to previous studies, we will obtain information from three groups: men with biopsy confirmed prostate cancer, men with negative biopsies, and men with a normal PSA and no history of prostate conditions. The goals of this study are to determine whether circulating blood folic acid concentration is associated with measures of reported dietary intake of folate, and to determine whether blood folic acid concentration is associated with prostate cancer after accounting for known risk factors. Specific Aims Primary Aim #1: To evaluate the association between blood folic acid concentration and dietary folate consumption as measured by a food frequency questionnaire. o Hypothesis for Primary Aim #2: Blood folic acid concentration will be significantly correlated with dietary folate consumption. Primary Aim #2: To evaluate the association between blood folic acid concentration and prostate cancer. o Hypothesis for Primary Aim #1: Those with high blood folic acid concentration and those with low blood folic acid concentration will have differing odds of having prostate cancer 5

14 METHODS Overview This case control study added blood folic acid data to existing data collected for a Diet and Prostate Cancer (DPC) study conducted between February 2002 and June The goal of the DPC study was to examine dietary influences on prostate cancer in combination with known risk factors 25. The goal of this study was to determine whether blood folic acid concentration is associated with prostate cancer risk and to what extent blood folic acid concentration correlates with reported dietary intake. Finally, we built a multivariate model to assess the relationship between prostate cancer and the existing dietary and risk factor data in combination with blood folic acid data. Recruitment and Informed Consent Identification and selection of subjects has been described previously 25. Briefly, cases and biopsy negative controls included men with an elevated prostate specific antigen (PSA) level who were referred by their primary care physician to the VA urologic oncology (UO) clinic for a prostate biopsy. Clinic controls were identified as those men >50 years of age who had an upcoming appointment at the primary care clinic at the VA and who had a normal PSA reading within the previous year, no history of prostate conditions, and were not currently being treated for prostate conditions, dementia, or existing cancer. Research coordinators contacted each man to determine whether he would be willing to participate. Participants were asked to come to the VA prior to scheduled biopsy or primary care appointment to complete the Informed Consent process and to 6

15 provide demographic and risk factor information, complete the Diet History Questionnaire, and provide a blood sample. For this analysis, men who were diagnosed with prostatic intraepithelial neoplasia (PIN) as a result of their biopsy were excluded due to the known status of PIN as a precurser to prostate cancer (n = 36). Men who did not provide a blood sample for folic acid analysis (n = 124) and men who did not complete the dietary and risk factor questionnaires (n = 11) were also excluded, leaving a total of 537 subjects (120 cases, 236 biopsy negatives, and 181 PSA normal controls) for this analysis. Details of those subjects who were identified and those who were included in this analysis are shown in Figure 1. All participants provided written informed consent according to both the PVAMC and OHSU Institutional Review Boards requirements and approval. 7

16 PSA normal controls n=238 Successful Interview n=732 (58%) Biopsy n=431 Total identified n=1569 Contacted n=1256 (81%) Unable to contact n=293 (19%) Refusal n=320 (25%) No show/other n=204 (16%) No biopsy n=63 PSA normal controls n=238 Provided blood sample n= 190 (80%) Completed DHQ n= 188 (99%) Total Subjects n=669 Prostate cancer n=140 (32%) Provided blood sample n=121 (86%) Completed DHQ n=120 (99%) Biopsy n=431 PIN n=36 (8%) Biopsy negative n=255 (59%) Provided blood sample n=248 (97%) Completed DHQ n= 244 (98%) Figure 1. Sample population Measurement of Predictor Variables Diet and Other Risk Factors Information on diet history was obtained using an adapted version of the National Cancer Institute Diet History Questionnaire (DHQ) 26. Participants were asked to recall their usual dietary intake of 124 foods, nutrient supplement use, beverage consumption and use of herbal remedies for the previous 12 months. Intake of individual nutrients was determined by analysis of the DHQ through the NCI Diet*Calc 8

17 software. Intake of folate and folic acid was calculated in several ways. First, natural folate from food sources was captured as Food Folate. Second, synthetic folate from fortified foods was captured as Synthetic Folate. Supplemental Folic Acid was calculated as folic acid coming from vitamin supplements. Finally, because food folate has lower bioavailability than synthetic folate 27, a Dietary Folate Equivalents variable was calculated as the sum of Food Folate and (Synthetic Folate multiplied by 1.7). The risk factor questionnaire requested information on age, race, ethnic origin, body mass index (BMI), education, family history of cancer, use of non steroidal anti inflammatory drugs (NSAIDs), supplement use, comorbid conditions, occupational history, alcohol consumption and smoking. Blood Folic acid Analysis A blood specimen was drawn from each participant using a vacutainer tube. Tubes were stored at 5 10 C in the dark no longer than 5 hours before processing. 0.1mL of whole blood was aliquoted to each of two cryovials and 0.5mL freshly prepared 0.8% sodium ascorbate was added to hemolyze red blood cells and stabilize folic acid. Tubes were inverted gently to mix and allowed to stand at room temperature (22 27 C) for complete hemolysis. Cryovials were then frozen at 80 C until analysis. All samples were prepared and analyzed at the General Clinical Research Center at OHSU. Samples were analyzed in batches containing similar numbers of cases and each of the two control groups by laboratory personnel blinded to the case control status of the specimens. Folic acid concentrations were determined using a competitive, ligandlabeled, protein binding chemiluminescent assay on an IMMULITE 1000 (Siemens 9

18 Healthcare Diagnostics, Deerfield, IL). Control samples (Lyphochek Whole Blood Control, Bio Rad Laboratories, Hercules, CA) were run between every 60 samples. Intraassay and interassay CVs were 6.6% and 12.1%, respectively. Measurement of Outcome Variable Presence of prostate cancer was determined via histological examination of biopsy tissue. Severity of disease was characterized by TNM staging and Gleason score. Statistical Analysis The objectives of the current study were to (1) determine the distribution of whole blood folic acid concentration in our sample, (2) assess the homogeneity of the distribution of folic acid concentration in cases and controls, (3) evaluate the association between blood folic acid concentration and reported dietary intake of folate and folic acid, and (4) build a multivariate model using laboratory, dietary and risk factor data to predict the outcome of prostate cancer. All statistical analyses were performed using the Statistical Analysis System (SAS) program, version 9.1 (SAS Institute, Inc., Cary, North Carolina). First, differences in demographic and risk factor information were evaluated between subjects who did and did not provide a blood sample to assess the presence of bias. Second, correlations were assessed between blood folic acid concentration and all measures of reported dietary intake including food folate, synthetic folate, supplementary folic acid, and dietary folate equivalents. Intake of the different measures of folate and circulating concentration of folic acid were categorized into quartiles based on the distribution in the entire sample for logistic regression analysis. 10

19 Dietary and other risk factor and demographic information were used to build multivariate logistic regression models comparing prostate cancer cases with either biopsy negative subjects or clinic control subjects. First, univariate analyses were conducted on the association of prostate cancer outcome with known and possible risk factors (age, race, family history, BMI, smoking), blood folic acid concentration, and dietary factors (alcohol consumption, food folate, synthetic folate, supplementary folic acid, dietary folate equivalents, and vitamin B6 intake). Those variables found to have a significant (p 0.25) association with prostate cancer were retained in the initial multivariate model. Variables that were not significant in the initial multivariate model were excluded individually to determine whether their absence significantly (>10%) changed the beta estimates of remaining predictor variables. Had any excluded variable exhibited such a change, it would be retained in the model. To evaluate the effect of these potential risk factors on prostate cancer of different serverity, cases were grouped according to Gleason score (< 7 and 7). The methods described above were used to build multivariate logistic regression models comparing all of the PSA normal clinic controls with these two groups of cases. Due to the fixed sample size of this study, our power to detect significant differences in prostate cancer risk is limited. This study was originally designed to detect differences in fatty acid consumption and the sample size was calculated for this purpose. This study has power of 80% (alpha = 0.05) to detect relative risks for prostate cancer as close to unity as 0.40 and 2.46 between cases and biopsy negatives, and as 11

20 close to unity as 0.43 and 2.31 between cases and clinic controls in the highest and lowest quartiles of blood folic acid concentration. 12

21 RESULTS Demographics Because a number of subjects did not provide blood samples, potential demographic differences between those who did and did not provide blood samples were evaluated and are shown in Table 2. No significant differences were found in age, case/control status, race, BMI, smoking status, family history of prostate cancer, or education. Demographic characteristics of cases and controls who provided both blood samples and dietary data, listed by subject group, are shown in Table 3. Mean ages for all three groups were similar, but the biopsy negative group had a significantly greater proportion of those under 55 years of age than the cases and clinic controls. All three subject groups were predominantly White, reflecting the racial makeup of Portland area residents, although the biopsy negative group had a greater proportion of both Asian and Native American/Alaska Native subjects than the other subject groups. Clinic controls were significantly heavier than cases, with a greater proportion of subjects with BMI >35. Clinic controls were also less likely to have a family history of prostate cancer. Biopsy negative subjects were more likely to have never smoked and also had more education than cases. Folate Next, the distribution of blood folic acid concentration was examined. The overall mean concentration was 839 ng/ml, and the distribution was strongly rightskewed with a median concentration of 786 ng/ml and a range of 99 ng/ml to

22 ng/ml. This distribution did not vary significantly by subject group. Among cases, the mean folic acid concentration was 824 ng/ml, the median was 792 ng/ml, and the range was ng/ml. Among biopsy negative controls, the mean folic acid concentration was 846 ng/ml, the median was 783 ng/ml, and the range was ng/ml. Among clinic controls, the mean folic acid concentration was 847 ng/ml, the median was 797 ng/ml, and the range was ng/ml. Histograms showing the distribution of blood folic acid concentration among cases, biopsy negative controls, and clinic controls can be found in Figures 2, 3, and 4, respectively. These figures indicate that the distribution of blood folic acid concentration is fairly similar across subject groups, although the biopsy negative controls had more outlying observations than the other groups. Next I assessed the distribution of all measures of folate status (blood folic acid concentration, food folate intake, synthetic folate intake, Dietary Folate Equivalents, and use of supplemental folic acid) among the three subject groups (Table 4). None of the distributions were significantly different between cases and either of the two control groups, although the distribution of Dietary Folate Equivalents approached significance when compared between cases and clinic controls (X 2 3 = 7.60, p = 0.055). Next I assessed the association between blood folic acid concentration and the different measures of dietary folate intake as captured by the DHQ. Correlation coefficients were low to moderate and ranged from for food folate to 0.34 for supplemental folic acid (Table 5). 14

23 Table 2. Characteristics of Subjects Who Did and Did Not Provide Blood Samples Characteristic Provided Blood Sample Did Not Provide Sample (n=544) (n=124) p value Age, in Years Mean Age Range N % N % Subject Group 0.31 Clinic Control Biopsy Negative Prostate Cancer Ethnic Origin 0.88 White Black Hispanic Native American Asian Other/Missing BMI 0.82 Less Than Family History of Prostate Cancer 0.84 Yes Smoking Status 0.95 Never Former Current Missing Education years Some College/Tech College graduate Missing/Other

24 TABLE 3. Frequencies of Demographic and Risk Factor Characteristics Characteristic Prostate cancer Biopsy negative controls PSA normal clinic cases (n=120) (n=236) controls (n=181) N % N % p value* N % p value* Age, in years Ethnic Origin White Black Hispanic Native American Asian Other/Missing BMI Family history of prostate cancer Yes Smoking Never Former Current Missing Education years Some College/Tech College graduate Missing/Other *P value for chi square difference between cases and respective control groups 16

25 TABLE 4. Distribution of Measures of Folate Status Characteristic Prostate cancer Biopsy negative PSA Normal Controls cases (n=120) controls (n=236) (n=181) N % N % p value* N % p value* Blood folic acid concentration ng/ml ng/ml ng/ml > 1098 ng/ml Food folate intake µg µg µg > µg Synthetic folate intake µg µg µg > µg Dietary Folate Equivalents µg µg µg > µg Use of Supplemental Folic Acid *P value for chi square difference between cases and respective control groups 17

26 Figure 2. Distribution of blood folic acid concentration among prostate cancer cases P e r c e n t Blood folic acid(ng/ml) Figure 3. Distribution of blood folic acid concentration among biopsy negative controls P e r c e n t Blood folic acid (ng/ml) 18

27 Figure 4. Distribution of blood folic acid concentration among clinic controls P e r c e n t Blood folic acid (ng/ml) Table 5. Correlation coefficients and corresponding p values for measures of folate status Blood folic acid Food folate Blood folic acid Food folate Synthetic folate Dietary Folate Equivalents Synthetic folate < < <.0001 Dietary Folate Equivalents < < <.0001 Supplemental folic acid < Total folate intake < < < <.0001 Supplemental folic acid < <.0001 Total folate intake < < < < <

28 Logistic Regression Modeling For logistic regression modeling I ran two separate analyses comparing prostate cancer cases with biopsy negative subjects and prostate cancer cases with clinic controls. I first examined univariate models for each variable of interest, including blood folic acid concentration, dietary folate intake (food folate, synthetic folate and dietary folate equivalents), use of supplemental folic acid, vitamin B6 intake, age, race, family history of prostate cancer, education, alcohol consumption, BMI, smoking, and PSA (for cases compared to biopsy negative subjects only). Models examining categorical variables used design variables to assess the odds of having prostate cancer in each category compared to a reference category. Univariate data for dietary factors are shown in Table 6, and data for other risk factors are shown in Table 7. Those variables found to be significant at the p <0.25 level were included in the initial multivariate model. Variables that were not significant in the initial multivariate model were excluded individually to assess the impact on the remaining predictor variables. Multivariate models were adjusted for Vitamin B6 intake due to its biologic interaction with folate. 20

29 Table 6. Univariate comparisons of dietary data between PCa cases and both control groups Cases vs. Clinic Controls Cases vs. Biopsy Negative Variable Crude OR (95% CI) p value Crude OR (95% CI) p value Blood folic acid concentration ng/ml Referent Referent ng/ml 1.11 (0.57, 2.13) 0.78 (0.42, 1.45) ng/ml 0.92 (0.48, 1.8) 0.73 (0.39, 1.38) > 1098 ng/ml 1.22 (0.65, 2.29) 1.15 (0.62, 2.12) Dietary folate equivalents µg Referent Referent µg 0.77 (0.40, 1.47) 1.06 (0.59, 1.91) µg 0.43 (0.22, 0.84) 0.84 (0.45, 1.57) > µg 0.52 (0.26, 1.01) 0.72 (0.39, 1.33) Food folate µg Referent Referent µg 0.92 (0.47, 1.79) 0.76 (0.41, 1.4) µg 0.84 (0.44, 1.58) 1.17 (0.64, 2.16) > µg 0.61 (0.32, 1.18) 0.87 (0.47, 1.64) Synthetic folate µg Referent Referent µg 0.69 (0.36, 1.32) 0.71 (0.39, 1.32) µg 0.6 (0.32, 1.15) 0.76 (0.41, 1.42) > µg 0.78 (0.41, 1.48) 0.82 (0.44, 1.47) Use of supplementary 1.22 (0.76, 1.97) (0.75, 1.86) 0.46 folic acid Vitamin B6 intake (mg) 0.85 (0.63, 1.13) (0.75, 1.25) 0.78 Alcohol (drinks per day) 1.15 (0.99, 1.34) (0.96, 1.14)

30 Table 7. Additional univariate comparisons between PCa cases and two control groups Cases vs. Clinic Controls Cases vs. Biopsy Negative Variable Crude OR (95% CI) p value Crude OR (95% CI) p value Age Referent Referent (0.61, 4.53) 2.02 (0.79, 5.19) (0.56, 4.29) 2.27 (0.87, 5.9) > (1.01, 12.88) (3.37, 58.83) Race White Referent Referent Black 3.08 (0.91, 10.49) 2.58 (0.87, 7.63) Other 0.64 (0.22, 1.88) 0.42 (0.16, 1.14) Family history of PCa 2.44 (1.06, 5.64) (0.49, 1.83) 0.89 PSA 0.11 <4 Referent (0.65, 2.2) > (0.99, 4.64) BMI < 25 Referent Referent (0.45, 1.76) 0.93 (0.5, 1.75) (0.44, 1.77) 0.95 (0.5, 1.79) (0.15, 0.82) 0.77 (0.34, 1.75) Smoking status Never Referent Referent Former 1.47 (0.76, 2.83) 2.16 (1.17, 3.98) Current 1.48 (0.71, 3.07) 2.64 (1.32, 5.3) Education High School Referent Referent Some/Technical College 0.73 (0.42, 1.25) 0.44 (0.26, 0.74) College Graduate 0.72 (0.4, 1.29) 0.45 (0.26, 0.8) Prostate Cancer Cases Compared to Clinic Controls In the comparison of cases with clinic controls, age, race, BMI, family history of prostate cancer, and three different measures of folate intake (food folate, synthetic folate, and dietary folate equivalents) were found to be significantly associated with prostate cancer at the univariate level, and all were included in the multivariate model. Because dietary folate equivalents comprises both food and synthetic folate, separate 22

31 models were built and included either dietary folate or both food and synthetic folate. Blood folic acid concentration was also included in the multivariate model despite its lack of significance at the univariate level. Among the dietary folate variables, only dietary folate equivalents remained significant in the multivariate model, so the model including this variable was pursued. Age, BMI, and family history of prostate cancer remained significant in the multivariate model whereas race and blood folic acid concentration did not. Race and blood folic acid concentration were retained in the model due to the status of race as a known risk factor for prostate cancer and the focus on blood folic acid concentration in this study. Vitamin B6 was also not significant but was retained to provide adjustment related to the folate measures. Subjects with both the highest and second highest quartile of intake of dietary folate equivalents were less likely to have prostate cancer, with odds ratios of 0.28 and 0.26, respectively, compared to the lowest quartile of folate consumption (95% CI and , respectively). Those with BMI >35 were also less likely to have prostate cancer, with an odds ratio of 0.31 compared to BMI <25 (95% CI ). Having a family history of prostate cancer also increased the likelihood of having prostate cancer with an odds ratio of 2.55 (95% CI ). This model does not exhibit significant lack of fit according to the Hosmer Lemeshow goodness of fit test (X 2 8 = 10.44, p = 0.23). Details of this model are shown in Table 8. 23

32 Table 8. Final multivariate model comparing prostate cancer cases (n= 120) with clinic controls (n=181) Variable Crude OR (95% CI) Adjusted OR (95% CI) p value Blood folic acid concentration ng/ml Referent Referent ng/ml 1.11 (0.57, 2.13) 1.57 (0.77, 3.22) ng/ml 0.92 (0.48, 1.8) 1.16 (0.56, 2.39) > 1098 ng/ml 1.22 (0.65, 2.29) 1.46 (0.73, 2.92) Dietary folate equivalents µg Referent Referent µg 0.77 (0.40, 1.47) 0.51 (0.24, 1.08) µg 0.43 (0.22, 0.84) 0.28 (0.12, 0.64) > µg 0.52 (0.26, 1.01) 0.26 (0.084, 0.78) Age Referent Referent (0.61, 4.53) 2.11 (0.7, 6.33) (0.56, 4.29) 1.94 (0.64, 5.94) > (1.01, 12.88) 5.6 (1.38, 22.67) Race 0.19 White Referent Referent Black 3.08 (0.91, 10.49) 3.39 (0.9, 12.79) Other 0.64 (0.22, 1.88) 0.93 (0.29, 3.01) BMI 0.07 < 25 Referent Referent (0.45, 1.76) 0.71 (0.35, 1.47) (0.44, 1.77) 0.79 (0.38, 1.64) (0.15, 0.82) 0.31 (0.13, 0.78) Family history of PCa 2.44 (1.06, 5.64) 2.55 (1.04, 6.24) 0.04 Vitamin B6 intake (mg) 0.85 (0.63, 1.13) 1.33 (0.83, 2.13) 0.24 Prostate Cancer Cases Compared to Biopsy Negative Controls In the comparison of cases with biopsy negative subjects, age, race, family history of prostate cancer, PSA, smoking, and education were found to be significantly associated with prostate cancer at the univariate level. All of these variables were included in the initial multivariate models along with blood folic acid concentration. Although no measure of dietary folate intake was significant at the univariate level, 24

33 because of the focus of this study on folate an initial multivariate model was built to contain dietary folate equivalents (Table 9). Table 9. Initial multivariate model comparing prostate cancer cases with biopsy negative controls Variable Crude OR (95% CI) Adjusted OR (95% CI) p value Blood folic acid concentration ng/ml Referent Referent ng/ml 0.78 (0.42, 1.45) 0.91 (0.45, 1.83) ng/ml 0.73 (0.39, 1.38) 0.85 (0.41, 1.73) > 1098 ng/ml 1.15 (0.62, 2.12) 1.27 (0.62, 2.59) Age Referent Referent (0.79, 5.19) 2.72 (0.94, 7.86) (0.87, 5.9) 2.9 (0.97, 8.64) > (3.37, 58.83) (3.58, 85.65) Race 0.07 White Referent Referent Black 2.58 (0.87, 7.63) 3.44 (1.0, 11.89) Other 0.42, 0.16, 1.14) 0.55 (0.19, 1.61) Family History of PCa 0.95 (0.49, 1.83) 1.08 (0.53, 2.22) 0.83 PSA 0.55 < 4 Referent Referent (0.65, 2.2) 1.14 (0.59, 2.2) > (0.99, 4.64) 1.58 (0.67, 3.73) Smoking status 0.04 Never Referent Referent Former 2.16 (1.17, 3.98) 2.01 (1.04, 3.89) Current 2.64 (1.32, 5.3) 2.62 (1.21, 5.71) Education 0.08 High School Referent Referent Some/Technical College 0.44 (0.26, 0.74) 0.59 (0.32, 1.11) College Graduate 0.45 (0.26, 0.8) 0.59 (0.3, 0.94) Vitamin B6 intake (mg) (0.75, 1.25) 1.31 (0.83, 2.07) 0.24 Dietary folate equivalents µg Referent Referent µg 1.06 (0.59, 1.91) 0.76 (0.38, 1.54) µg 0.84 (0.45, 1.57) 0.59 (0.26, 1.32) > µg 0.72 (0.39, 1.33) 0.43 (0.14, 1.27) 25

34 In the initial multivariate model, dietary folate equivalents was not significantly associated with prostate cancer and was excluded from the final model. Vitamin B6 intake was retained to provide adjustment relative to blood folic acid concentration, although neither of these variables was significant in the multivariate models. All other variables remained significant except family history and PSA; PSA was eliminated whereas family history was retained due to its clinical significance. The final model included blood folic acid concentration, age, race, smoking, education, and vitamin B6 (Table 10). Subjects in the three highest age categories were more likely to have prostate cancer than those under 55 years of age, although the association was significant only for the comparison of those greater than 75 years compared to those under 55 years (OR= 17.26, 95% CI ). Black subjects were more likely to have prostate cancer than white subjects with an odds ratio of 3.66 (95% CI ). Both former and current smokers were more likely to have prostate cancer than never smokers with odds ratios of 1.96 and 2.82, respectively (95% CI and , respectively). Finally, subjects who attended some college or technical college and those who were college graduates were less likely to have prostate cancer than those with a high school education or less, although the association was significant only for those with some college/technical college education (OR = 0.52, 95% CI ). This model does not exhibit significant lack of fit according to the Hosmer Lemeshow goodness of fit test (X 2 8 = 2.94, p = 0.94). 26

35 Table 10. Final multivariate model comparing prostate cancer cases with biopsy negative controls Variable Crude OR (95% CI) Adjusted OR (95% CI) p value Blood folic acid concentration ng/ml Referent Referent ng/ml 0.78 (0.42, 1.45) 0.9 (0.46, 1.8) ng/ml 0.73 (0.39, 1.38) 0.8 (0.4, 1.62) > 1098 ng/ml 1.15 (0.62, 2.12) 1.2 (0.59, 2.41) Age Referent Referent (0.79, 5.19) 2.62 (0.92, 7.44) (0.87, 5.9) 2.84 (0.97, 8.31) > (3.37, 58.83) (3.66, 81.33) Race 0.04 White Referent Referent Black 2.58 (0.87, 7.63) 3.66 (1.09, 12.25) Other 0.42, 0.16, 1.14) 0.5 (0.18, 1.43) Family History of PCa 0.95 (0.49, 1.83) 1.11 (0.54, 2.26) 0.78 Smoking status 0.03 Never Referent Referent Former 2.16 (1.17, 3.98) 1.96 (1.02, 3.76) Current 2.64 (1.32, 5.3) 2.82 (1.31, 6.05) Education 0.05 High School Referent Referent Some/Technical College 0.44 (0.26, 0.74) 0.52 (0.3, 0.91) College Graduate 0.45 (0.26, 0.8) 0.55 (0.3, 1.02) Vitamin B6 intake (mg) (0.75, 1.25) 0.99 (0.74, 1.31)

36 Stratification by Gleason Score Because risk factors for prostate cancer may vary according to severity of disease, I stratified the comparison of cases with clinic controls according to Gleason score (< 7, n=57, and 7, n=60). Gleason score was not available for three cases; these subjects were excluded from this analysis. A Gleason score of 7 or greater conveys an increased risk of metastasis and recurrence 28. Stratification of the results of this study indicated that risk factors may indeed vary according to disease severity. Among men with Gleason 7, age, race, BMI, and family history of prostate cancer were significant predictors of prostate cancer (Table 11). In contrast, among men with Gleason < 7, only total drinks of alcohol was a significant predictor whereas the known risk factors of age, race, and family history were not (Table 12). In addition, although Dietary Folate Equivalents was significant in the univariate analysis, this association was no longer significant after adjustment for vitamin B6. Blood folic acid concentration was not associated with either category of prostate cancer severity but was retained in both models. As with the other multivariate models, vitamin B6 intake was included to provide adjustment relative to the blood folic acid measure. 28

37 Table 11. Multivariate model comparing severe prostate cancer cases (Gleason 7, n=60) with clinic controls (n=181) Variable Crude OR (95% CI) p value Adjusted OR (95% CI) p value Blood folic acid ng/ml Referent Referent ng/ml 1.3 (0.58, 2.91) (0.8, 4.99) ng/ml 0.85 (0.36, 2.02) 1.08 (0.42, 2.79) > 1098 ng/ml 1.11 (0.49, 2.51) 1.13 (0.46, 2.8) Age Referent Referent (0.5, 10.96) 2.88 (0.49, 16.79) (0.5, 10.96) 2.95 (0.5, 17.31) > (1.32, 42.76) 10.3 (1.43, 74.14) Race White Referent Referent Black 1.21 (0.33, 4.47) 7.1 (1.78, 28.3) Other 6.99 (1.2, 40.82) 0.9 (0.21, 3.78) BMI < 25 Referent Referent (0.28, 1.45) 0.49 (0.2, 1.19) (0.37, 1.88) 0.69 (0.29, 1.67) (0.05, 0.6) 0.15 (0.04, 0.56) Family history of PCa 2.63 (0.99, 7.01) (1.19, 10.78) Vitamin B6 intake (mg) 0.95 (0.66, 1.38) (0.68, 1.58)

38 Table 12. Multivariate model comparing less severe prostate cancer cases (Gleason < 7, n=57) with clinic controls (n=181) Variable Crude OR (95% CI) p value Adjusted OR (95% CI) p value Blood folic acid ng/ml Referent Referent ng/ml 0.84 (0.35, 2.02) 1.06 (0.42, 2.65) ng/ml 1.0 (0.43, 2.29) 1.13 (0.45, 2.83) > 1098 ng/ml 1.18 (0.53, 2.64) 1.38 (0.59, 3.21) Dietary folate equivalents µg Referent µg 0.57 (0.25, 1.29) µg 0.31 (0.13, 0.75) > µg 0.48 (0.21, 1.1) Age Referent Referent (0.39, 4.13) 1.31 (0.36, 4.74) (0.32, 3.55) 1.03 (0.27, 3.86) > (0.33, 8.37) 1.68 (0.3, 9.34) Race White Referent Referent Black 1.96 (0.43, 9.05) 0.79 (0.08,. 8.22) Other 1.5 (0.11, 21.31) 0.77 (0.16, 3.8) Smoking Status 0.15 Never Referent Former 2.71 (0.99, 7.45) Current 2.13 (0.70, 6.48) Alcohol (drinks per day) 1.24 (1.04, 1.49) (1.04, 1.51) Family history of PCa 2.39 (0.87, 6.61) (0.71, 6.48) 0.18 Vitamin B6 intake (mg) 0.78 (0.53, 1.13) (0.45, 1.03)

39 DISCUSSION Although these data showed no significant association between blood folic acid concentration and prostate cancer, the research question remains important and is worthy of continued future study. In addition, the results of this study lead to intriguing questions about associations between the variables examined. Blood folic acid concentration Blood folic acid concentration was not significantly associated with prostate cancer in either the comparison of cases with clinic controls or the comparison of cases with biopsy negative controls. There are several possible explanations for this observation. First, the null association could be true. Second, the association could exist, but have smaller magnitude than we had power to detect, given the sample size of this study. Third, the assay we used to measure blood folic acid concentration could have been inaccurate or inadequately sensitive. Finally, given recent discussions on the possibility of differential action of folate depending on the timing of exposure, it is possible that the association between blood folic acid concentration and prostate cancer may only be evident with measurements preceding carcinogenesis 29. The third explanation could be plausible for a few reasons. Red blood cell folic acid concentration is an accurate measure of body stores 30, however we were able to measure whole blood folic acid only, as the hematocrit of the sample was not measured. In addition, we measured only one replicate of each sample, and given the variation seen in control samples (intraassay and interassay CVs were 6.6% and 12.1%, respectively) additional replicates would have improved the accuracy of the 31

40 measurements. Furthermore, it is unclear whether the folic acid binding protein (FBP) used in the assay has the same binding affinity for all forms of folate that may be present in the sample. It is known that some FBPs exhibit higher affinity for oxidized forms of folic acid than for reduced folate 31. The Immulite assay used in this study includes a pre treatment step with dithiothreitol (DTT), a reducing agent, to ensure similar binding affinities for folate and folic acid, however I was unable to verify binding affinities for this particular FBP. It is not surprising that blood folic acid concentration was not highly correlated with any of the dietary intake measures when evaluated as continuous or categorical variables (quartiles). One other study of folate and prostate cancer reported low correlation between serum folate and reported intake among participants with a correlation coefficient of One study of NHANES data also reported low correlation between reported dietary intake of folate and both serum and RBC concentrations; reported correlation coefficients ranged from 0.11 to 0.29 in different subpopulations 32. This is likely due in part to the different ways questionnaires and biochemical assays measure folate. Questionnaires may ask for reported intake during the previous day, week, or year, and may be given at a different time point than the blood or serum sample is taken. Consistency in the measurement of both dietary and biochemical indicators of folate status would improve these disparities. Recent studies have suggested that natural and supplemental folic acid may act differently with regard to prostate cancer risk. In two randomized controlled trials, supplemental folic acid was shown to increase risk of both colorectal and prostate 32

41 cancer whereas baseline dietary intake of folate was associated with a decreased risk of prostate cancer in the Aspirin/Folate Polyp Study 8,23. In addition, recent analysis of the United States and Canada, both of which have folic acid fortification programs, has found an increase in colorectal cancer prevalence since the fortification programs began 33. Although the data in this study do not support a positive association between folic acid and prostate cancer, those studies that do report such an association should not be ignored. It is clear that further studies are necessary to elucidate the effects of both dietary folate and folic acid on cancer risk. Comparison of Cases and Clinic Controls In the multivariate model comparing prostate cancer cases with clinic controls, Dietary Folate Equivalents (DFE), BMI, and family history of prostate cancer were significantly associated with prostate cancer, whereas blood folic acid concentration, age and race were not. Subjects with both the highest and second highest quartile of DFE were less likely to have prostate cancer than those with the lowest quartile of DFE. It is interesting that DFE was significantly associated with prostate cancer whereas the components of DFE, food folate and synthetic folate, were not significant. This suggests that although both food and synthetic folate may contribute to prostate cancer risk, the combination of these two sources of folate, taking into account their different bioavailability, is a stronger contributor than either of the individual components. Increasing BMI was associated with a reduced odds of having prostate cancer. This association can be attributed to the difference in proportions of subjects in the highest category of BMI (>35) in cases and clinic controls. 23.8% of clinic controls had 33